INDEX

Explanations

keywords related to summarizing and writing headlines

oai_token-act-pair · gpt-3.5-turbo

prominent noun phrases and introductory phrases in the text

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

EngineDebug

-0.52

ãĥ¼ãĥ³

-0.48

arij

-0.47

ãĥ¼ãĥĨ

-0.46

chuk

-0.45

ä»£

-0.45

 Moroc

-0.45

],"

-0.45

ãĥ¼ãĥĨãĤ£

-0.44

POSITIVE LOGITS

cknowled

0.55

cing

0.49

 largeDownload

0.47

terness

0.43

othes

0.41

 DRAGON

0.40

 hurry

0.39

 comforting

0.38

ITS

0.38

 mysterious

0.37

Activations Density 0.645%

keywords related to summarizing and writing headlines

prominent noun phrases and introductory phrases in the text

No Comments

No Known Activations